The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes remote commands via ssh and utilizes sudo for administrative system tasks such as package management and device configuration via raspi-config.
[CREDENTIALS_UNSAFE]: The skill manages sensitive SSH private keys using the $RASPI_SSH_KEY environment variable and provides instructions for modifying file permissions (chmod 600) to manage these credentials.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs software from external repositories using apt and pip, specifically for tools like apache2, mosquitto, and the RPi.GPIO library.
[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection by monitoring external system data such as the Apache access logs. * Ingestion points: System and web server logs (/var/log/apache2/access.log). * Boundary markers: No specific delimiters or safety warnings are implemented to prevent the agent from obeying instructions embedded in the logs. * Capability inventory: Includes remote shell access via SSH, privileged command execution via sudo, and script generation. * Sanitization: Log content is processed without validation or sanitization.

raspberry-pi